Tag
This paper proposes a quantization-robust unlearning framework for large language models, using loss landscape analysis to ensure effective forgetting while maintaining model utility after compression.